VSPAero: Panel Method Crashing

[Bhumika Dutta]

Hi,

I'm a beginner to OpenVSP, VSPAero and I'm trying to use the panel method to analyze a certain aircraft configuration. I just want to be able to go through the results and look at how various graphs are behaving. However, whenever I try launching solver, the software crashes. Here's what I'm inputting when I try to run the panel method:

Geometry Set: WingBody (I'm not sure what Geometry Set is referring to)

From Model, Ref.Wing: Mainwing 

Mach Start: 0.2 

Let me know if more info is needed. I've just been playing around with the different parameters, if anyone can guide me to some kind of tutorial or documentation that describes what each parameter does, that would be great! I appreciate any help I can get!

[Rob McDonald]

I'd be willing to bet that you're trying to run in VLM mode and have a fuselage with rounded rectangle XSecs.  Unfortunately, that causes problems.  For VLM, you might as well change those XSecs to be ellipses.

If I haven't gotten a lucky guess, you're going to need to provide more information.

Rob

[Brandon Litherland]

With Panel method, blunt TE or any "vertical" aft-facing surface will crash the solver I believe due to the Kutta condition applied to trailing edges.  Either modify the airfoils to have sharp TE by going to Modify > TE > Closure > Skew Both or Skew Lower > 0.0 or if you are using native VSP airfoils make sure that Sharp TE is checked.

[Bhumika Dutta]

It's not crashing now, but I can't see anything in the results manager. What am I doing wrong? Also, is there a resource I could use to understand how VSPAero works, specifically the panel method since I don't want to use VLM. I appreciate the feedback!

Here's what I have: 

[Hemanth Sethuraman]

Dear Bhumika

The vsp workshop has most of the documents which explains how to use vsp aero. It should help you out. 

Many thanks 

Hemanth Kumar Sethuraman 

--
You received this message because you are subscribed to the Google Groups "OpenVSP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openvsp+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openvsp/1edd8678-4daf-478e-aa98-526ec4c8daf8n%40googlegroups.com.

[Brandon Litherland]

Why are you avoiding VLM?  I see you've got an Illinois email.  Are you with one of the design groups?

[Bhumika Dutta]

I'm working with the Center for High-Efficiency Electrical Technologies for Aircraft (CHEETA) at UIUC. We want to study the aerodynamics characteristics of this particular aircraft configuration prior to wind tunnel testing. We decided that the panel method would be better suited for our purposes. 

[C P]

Hi all,

looking a the second picture, the pannelling of al components require some work in order to match better the panels at intersections. Wings and horiz and vert stabs are to coarse, fuselage too fine at wing intersection. I often noticed that , although CompGeom is able to produce a watertight geometry, the solver in VSP has some problem to start. 

At the end, it is a work of trial and error, but it always worked for me in similar cases.

Corrado

[Brandon Litherland]

I did my undergrad work at UIUC and loved the program and faculty.  Lots of good memories from my time there.  One of my earliest tests of VSPAERO and OpenVSP was to see if I could get reasonable predictions for the Aero Lab course models in the small flowthrough tunnel.  Even back then, we got within 5% using VLM and parasite drag buildups.  I think that was 2014/2015 or so.

To your question specifically, Corrado is correct that having similar edge lengths along the junctions will help the mesh and solver.  You may also find that activating the new Experimental File Format option will improve the convergence and potentially the solution accuracy.  You'll also want to bump up the spanwise resolution of the wings and lifting surfaces.  The default of 6 is insufficient for VSPAERO.  I tend to start with at least 16 and move up to 21 or higher.  Anything more than 81 is probably overkill unless you're doing something like blown wings where there is a lot of flow interaction going on. You may also want to bump the chord resolution to something like 41 and adjust the LE/TE clustering to 0.2/1.0.  Be on the lookout in the Viewer for odd CP peaks particularly at the wingtip TE and caps.  These can mess with your convergence and final result.

There are some videos in the workshops as Hemanth mentioned.  The link will take you to the 2021 page were we have some VSPAERO best practices and some meshing guidance.  The 2020 Workshop has information more to do with operation of the solver from the GUI and command line.  And of course now I read your original question more closely and realize that you really are just getting into this.  Definitely check out the 2020 Workshop VSPAERO videos and presentations to get a feel for what's going on.  There is also a presentation on Attach, symmetry, sets, subsurfaces that will help you understand sets.  Basically a group of components that you want to include in an analysis or operation.

Quite a few of the CHEETA group should be very familiar with OpenVSP so keep asking around for guidance and you're always welcome to post questions here.  For fundamental OpenVSP instruction, go to the OpenVSP Ground Schoolwhere we have lots of quick tutorials explaining basic operations.  (I'm trying to get caught up on producing the rest of the content!)

- Brandon

[Bhumika Dutta]

Thanks so much for all the feedback! 

[Bhumika Dutta]

Hi,

I've started looking through the tutorials and made some of the adjustments mentioned. However, what do you mean when the wings, vert stabs are too coarse and the fuselage too fine at wing intersection? How do I improve it? Also, where do I change the chord resolution? I've made the changes to tessellation and clustering. Additionally, do I need to do something with CompGeom specifically before running panel method? 

Bhumika 

[Brandon Litherland]

The chord resolution just means the NumW setting.  When you have intersecting components, good practice is to try and keep the spacing between interpolated sections relatively similar so the intersections are well defined.  This will typically help to avoid any very slender "sliver" panels that can cause problems.  In the image you shared, for example, the aft section of the fuselage has large steps moving aftward but reasonably fine resolution/tessellation in the W direction around the body. Lots of skinny panels back there which should be avoided.  

You will not need to run CompGeom before running Panel.  It will do it for you and produce the mesh as a component.

[Bhumika Dutta]

Hi,

I've tried making the num w and u more consistent between each of the parts. I think the settings for each are all the same, however it's crashing again. I did notice, the mesh created is not including the main gear (selected in the red box) as well as the v tail. Why is that? 

Bhumika 

[Brandon Litherland]

If you run CompGeom or VSPAERO in panel mode, are the components included in the Set?

In the interest of time, can you upload this model to this thread so we can look at it on our machines?  There may be something going on that isn't clear in the images.

[Bhumika Dutta]

Here's the file. I don't believe that the component was included in the Set. 

[Rob McDonald]

I had no crash with your file.

I deleted the embedded MeshGeom.  I ran cases with the WingBody Set and also the All Set.

The sharp fuselage side towards the aft is detected as a trailing edge on which VSPAERO attaches a wake and applies the Kutta condition.  This may be appropriate or may not, depending on how you want to treat it.

The aft fuselage resolution (nose-to-tail direction) is very coarse -- it also has a great mismatch to the vee-tail chordwise resolution where they meet.  The blue hot spots may be caused by numeric issues due to this poor paneling, or may be related to the aerodynamics of the odd concave pocket at the root of the vee tail and the sharp chine edge of the aft fuselage -- they might be real consequences of an undesirable geometry.

I would recommend a slight increase in spanwise resolution on the main wing -- or at least some degree of clustering at the tip.

Rob

To view this discussion on the web visit https://groups.google.com/d/msgid/openvsp/82fb53b5-68ae-4ff2-968c-66e902289feen%40googlegroups.com.

[Brandon Litherland]

I'll follow up to Rob's responses.

I increased the U segments along the fuselage to get some more panels around the wing and the tails and to reduce the aspect ratio of some of the aft panels.  I also moved the tails forward in X just a bit to ensure that they were fully intersecting the body.  If you look closely you can see a little bit of the inboard root sticking out of the thin aft section.  You've also fallen victim to a zero strength skinning problem at your Point closeout section.  Zero strength skinning should be avoided at all costs.  Better practice is to deactivate the control on a given section that you just want to snap to the points.  Just uncheck all the boxes.

If you see something in the Viewer that doesn't look right, you can always check the triangulation, computational mesh, and agglomeration.  These options are available under the Aero menu or you can use the -/+ keys to cycle the agglom levels.  If you look at your intersections and see lots of jagged/skinny triangles where the mesh is having trouble stitching things together, that will usually indicate an area that needs work for the solver to resolve well.  Also, if you notice the solver running and running on a single GMRES iteration, up to 100+ steps or so, then it's chewing on something it doesn't like and you results may be affected.

[Bhumika Dutta]

Thanks, I will make changes. I was able to get it run now, and I'm seeing what Rob has now as well. What do you mean when you say that there's a zero strength skinning problem? Also, what is the Point closeout section? When I go to the skinning for the fuselage, I notice that the strength for skin cross sections 4 and 5 are set to 0. Is this what you are referring to?

[Brandon Litherland]

That's correct.  Zero strength skinning essentially collapses the nearest interpolated U segment onto the XSec.  If you turn on Shaded surfaces you will see a shadow or dark region where the skinning is set too low.  Even 0.01 will fix this but better practice is to deactivate the control.
The point closeout is the final XSec that makes the surfaces watertight.